Research map / Label-efficient and robust learning / Recognition and long-tail
Label-Free Supervision of Neural Networks with Physics and Domain Knowledge
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a fixed-curvature free-fall physics constraint instead of direct labels | CNN · Free-fall object tracking, test images | correlation: 12.1% -> 90.1% | Helped |
| Train directly on height labels rather than the physics constraint | CNN · Free-fall object tracking, test images | correlation: 90.1% -> 94.5% | Helped |
| Use constant-velocity constraints with nontrivial-output sufficiency terms rather than direct labels | CNN · Walking-person tracking, test set | correlation: 80.5% -> 95.4% | Helped |
| Remove the standard-deviation bonus h1 (ablation) | Walking-person constraint-learning method · Walking-person tracking | correlation | Hurt |
| Remove the output-range penalty h2 (ablation) | Walking-person constraint-learning method · Walking-person tracking | correlation | Hurt |
| Remove any one of h1, h2, or h3 (ablation) | Logical-constraint object detector · Peach and Mario detection | correct image descriptions | Hurt |
Closest papers by meaning
- A critical analysis of self-supervision, or what we can learn from a single image (ICLR 2020 · similarity 0.87)
- Beyond Learning Features: Training a Fully-Functional Classifier with ZERO Instance-Level Labels (AAAI 2022 · similarity 0.87)
- Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models (ICLR 2024 · similarity 0.86)
- Learning from weak labelers as constraints (ICLR 2025 · similarity 0.85)
- Deep Probabilistic Supervision for Image Classification (NeurIPS 2026 · similarity 0.84)
- Training the Untrainable: Introducing Inductive Bias via Representational Alignment (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.